A fault prediction method for generator bearings of offshore wind turbines based on DBO-XGBoost model and EWMA control chart
WU Qing
WANG Xiao
TAO Yanting
SONG Zeshuang
XU Linghua
YAN Jianguo
XING Xueshu
Abstract：[Objective]To detect faults in the generator bearings of offshore wind turbines in a timely manner,a DBO-XGBoost prediction model for predicting generator bearing temperature was proposed based on the dung beetle optimizer(DBO)algorithm and eXtreme Gradient Boosting(XGBoost)model.Combined with the exponentially weighted moving average(EWMA)control chart,this model enables fault prediction for generator bearings.[Methods]First,key features that can accurately characterize the operational status of generator bearings were selected from the supervisory control and data acquisition(SCADA)system by means of the maximal information coefficient(MIC),and these features were then fed into the DBO-XGBoost model to predict the generator bearing temperature under normal operating conditions.Second,the deviation between the actual values and the predicted values was quantified using the mahalanobis distance(MD),and the MD sequence was input into a change-point detection algorithm based on the exponentially weighted moving average(EWMA)control chart to identify the change points corresponding to fault occurrences,thereby enabling fault prediction.Finally,a knowledge graph for bearing fault modes was constructed on the basis of feature importance.[Results]The results demonstrate that this method can achieve relatively accurate prediction of generator bearing temperature under normal operating conditions and issue early fault warnings 3 days in advance.Compared with the fault warning method relying on setting a single threshold,this method can detect the time of fault occurrence with higher accuracy.In addition,the fault modes knowledge graph constructed for bearing provides visualized operation and maintenance decision support for operations personnel.
Keywords：Offshore wind turbineDung beetle optimization algorithmGenerator bearingFault predictionExponentially weighted moving average
Publication Date：2026-01-31
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：10( 133-142 )